Live Listen has been around since 2014. This decade-old accessibility feature is causing a stir for the second time this year.
I wonder how Siri Recap and Live Rewind will be received.
Side note: No, your iPhone isn’t listening to you. support.apple.com/guide/ipho…
This quoted post is unavailable.
Truly the end of an era. Probably great news for Reddit.
End of an era: Apple Support Community forums will permanently close on November 20 9to5mac.com/2026/11/10/end-o… by @mvcmendes
Apple may finally let Siri AI read Health data in iOS 27.3. Beta 1 ships with new, disabled feature flags.
IF/HealthSearch adds Health to Siri’s “find” tool. It can search recent readings and historical summaries across 92 HealthKit data types.
It also has dedicated queries for:
• Activity rings
• Individual workouts and workout summaries
• Readiness, and notably not Vitals
• Sleep summaries
• Medication schedules and logs
• Health Age and health assessments in the new Health app
IF/HealthSnippets allows Siri to use custom visualizations for:
• Health statistics
• Activity rings
• Individual workouts
• Workout summaries
• Sleep summaries
There’s no easy way to enable feature flags on platforms like iOS and watchOS.
ALT Two dark gray rounded-square app icons sit side by side on a black background. On the left, Siri AI shows a sphere that fades from dark gray at the top to light gray at the bottom, with a ribbon of iridescent curving across its middle. On the right, Health shows a glossy pink-red heart with a soft glow beneath it.
As of today, Apple notes that “Reading Health app data is not available with Siri AI at this time.” support.apple.com/en-us/1184…
1
15
1,559
Apple is collecting employee feedback to improve autocorrect.
A disabled iOS 27.3 flag, UIKit/keyboard_cohort, shows a dialog when you revisit an autocorrection.
Employees can file a Radar with intended text and diagnostics, sent to ISE User Research as a “Serious Bug.”
ALT iPhone 18 Pro showing a Messages draft to John that reads “I slept well last night.” An alert titled “Autocorrect got it wrong?” shows “well” changed to “we’ll” with options “File Radar” and “Keyboard Acted Correctly.”
To be clear: Siri AI’s Skimmer is a great fast path. It recognizes common queries and deterministically picks a tool and fills in its arguments without using an LLM.
This design makes Skimmer lightning fast. In some testing, it calls the “play” tool for music in under 120ms, up to 6× faster than the on-device or PCC model. It calls simpler tools like “flip coin” in under 60ms, up to 20× faster than models. Because no model runs, the speed is consistent across devices. It also works when offline!
And of course, if Skimmer can’t handle a request, Siri AI quickly compiles relevant context and sends your query to an on-device or PCC model.
Skimmer is the first step in Siri AI’s on-device processing. It uses regular expressions to match requests to tool flows.
On macOS, you can browse the rules by locale:
/System/Library/PrivateFrameworks/IntelligenceFlowPlannerSupport.framework/Versions/A/Resources/enigma_patterns/<locale>/Enigma.json
ALT Enigma.json file open in Xcode, showing the pattern and regular expression Skimmer uses to recognize named timer requests.
Skimmer is the first step in Siri AI’s on-device processing. It uses regular expressions to match requests to tool flows.
On macOS, you can browse the rules by locale:
/System/Library/PrivateFrameworks/IntelligenceFlowPlannerSupport.framework/Versions/A/Resources/enigma_patterns/<locale>/Enigma.json
ALT Enigma.json file open in Xcode, showing the pattern and regular expression Skimmer uses to recognize named timer requests.
Apple may be working on parking-aware ETAs.
iOS 27.3 adds a TimeToPark flag. When enabled, CoreRoutine uses background location and motion data to measure how long you spent near your destination before parking, and how long you walked afterwards. It saves both metrics per trip, grouped by day of the week and hour.
The feature can only be enabled on internal installs, and no Maps feature uses the data yet.
ALT A user’s preferred route to work is shown in Apple Maps on iPhone 16 Pro.
Apple continues to test Siri AI on all HomePod models.
In audioOS 27.3, the new HomePod descriptor grants the Siri AI capability to HomePod 2 and mini 2.
The two older models get the same capability when the Siri / linwood_enabled_for_candidate_homepods flag is enabled.
Here’s what’s going on with Siri AI on HomePod and Apple TV, and why HomePod 2 may end up getting it.
One of the prerequisites for enabling Siri AI on any device is a SiriMode hardware check through MobileGestalt. The hardware check relies on the device’s reported soc-generation: M10 or newer for HomePod.
Here are the soc-generations for each HomePod device:
• HomePod (OG): H7
• HomePod mini 1: M9
• HomePod (2nd generation): M10
• HomePod mini 2: M11
HomePod (2nd generation) and HomePod mini 2 pass the gate. However, it seems that Apple may have tested Siri AI on older hardware as well. If a HomePod device fails the generation check, it can qualify based on a defined list of product IDs. In audioOS 27.2, those include both HomePod (OG) and HomePod mini 1.
Why does this hardware gate matter? It turns out there’s an equivalent eligibility gate for all products. For Apple TV, it expects soc-generation H11 or newer. Here are the recent Apple TV 4K models:
• Apple TV 4K (1st generation): H9
• Apple TV 4K (2nd generation): H11
• Apple TV 4K (3rd generation): H14
• Apple TV 4K (4th generation): H18
Apple has confirmed that access to Siri AI is available on Apple TV 4K (2nd generation) or later. All three currently supported models pass the hardware check. There is no product ID bypass for this gate.
If both HomePod 2 and Apple TV 4K pass the hardware gate, why can’t you enable it today?
• On Apple TV, the Linwood feature is gated to internal builds, even if the hardware qualifies. Apple says Apple TV 4K (4th generation) gets Siri AI with tvOS 27.2, so we’ll see what ships.
• On HomePod, the Linwood feature explicitly returns false. This tracks with Apple’s “will be available later as a free software update” announcement.
On both products, there’s a Siri/linwood_override feature flag that can bypass the ordinary Linwood check.
As always, these gates may change before release.
ALT HomePod in two colors: white, midnight (black). Top exterior, flat backlit multi-color illuminated touch surface. Siri AI written above the speakers in orange and blue gradient.
Big news for virtualization: macOS 27.3 lets Core AI in Mac VMs use the host’s GPU and Neural Engine.
In some tests, the new paravirtualized paths can run Core AI tasks up to 7× faster¹ than the guest CPU, achieving 50 to nearly 100% of the host’s native speed.
The new Core AI accelerator currently requires private APIs and a restricted entitlement.
Here's how to set it up:
• Install macOS 27.3 on the host and guest.
• Build a VM host app with com\.apple.security.virtualization and com\.apple.private.virtualization entitlements. You’ll need to relax AMFI for the second entitlement.
• On VZVirtualMachineConfiguration, add the accelerators using _setAcceleratorDevices with:
• _VZMacCoreAIDeviceConfiguration (GPU, adds PVAIDevice)
• _VZMacNeuralEngineDeviceConfiguration (existing ANE accelerator, adds Neural Engine for Core AI)
On the guest, enable the feature flag:
sudo mkdir -p /Library/Preferences/FeatureFlags/Domain && \
sudo defaults write /Library/Preferences/FeatureFlags/Domain/ParavirtualizedCoreAI enable_pv_delegate -dict Enabled -bool true
and restart the guest. You can now call Core AI on the guest. If set up properly, ComputeUnitKind.availableKinds should list gpu and neuralEngine.
——
There's a second, currently disabled host-side feature flag (ParavirtualizedCoreAIHost/enable_save_restore). When enabled, the Core AI accelerator advertises support for saving and restoring VM state.
——
1. Tested on a Mac with an M4 Max chip using public models (YOLO, CLIP, Depth Anything) and matmul/convolution benchmarks. Crossing the VM boundary adds ~1 ms per inference call.
Qwen3-0.6B in Apple’s llm-benchmark ran at about 10% of host speed. Core AI’s paravirtualized path repeatedly sets up host access to large shared KV-cache buffers for each generated token.
ALT Core AI and Containerization icons side by side on a black background.
macOS 27.3 starts implementing changes to Full Disk Access (FDA). Here’s what ‘very explicit user action’ may mean.
When an app tries to read a protected file without FDA, macOS doesn’t prompt the user to grant permission. Instead, it automatically denies access and records that denial in the TCC database.
Once in the db, apps show up under Settings > Privacy & Security > Full Disk Access with their toggle off. Apps will often add themselves to the list, then instruct the user to open Settings and flip the toggle.
Not anymore: starting with macOS 27.3 and feature flag TCC/FDA_WARNING, automatic denials are no longer saved. Users will have to explicitly add an app to the list with the + button or by dragging it in from Applications.
So far, the flag doesn’t add any additional warnings or dialogs, but its name suggests these may arrive in future betas. The flag is disabled in 27.3 beta 1.
——
Relatedly, macOS 27.3 introduces “Private Folders,” a new folder-by-folder permission configured like Full Disk Access. Not fully wired up yet, but it could offer a narrower alternative to Full Disk Access.
It’s also impacted by TCC/FDA_WARNING: when the flag is on, automatic Private Folders denials aren’t saved either.
Mac apps will soon need ‘very explicit user action’ to gain full disk access, according to Apple. theverge.com/tech/1004295/ap…
Apple’s Passwords app gets a lock timeout control in macOS 27.3.
In macOS 27.2 or earlier, Passwords locks after 3 minutes of inactivity.
You can now choose a different timeout, or have the app lock when your Mac locks or sleeps, or never.
ALT Passwords Settings on macOS 27.3 with the “Automatically lock” menu open. Options are after 1, 3, or 10 minutes of inactivity, when the Mac locks or sleeps, or never. “After 3 Minutes of Inactivity” is selected.
macOS 27.3 brings controls for Suggestions in Messages Settings.
You can now pick what Suggested Actions you’d like to see underneath a message, or disable them all.
Also available in iOS 27.2 under Settings > Apps > Messages > Suggested Actions.
ALT Messages Settings on macOS 27.3 with a “Choose Suggestions” dialog. Available options are Add to Calendar, Add to Reminders, Add to Notes, Show in Maps, and Flights and Packages. All options are enabled. A note says suggestions you don’t use will be turned off after 30 days.
Apple is working to bring Siri AI to China.
OS 27.3 adds a disabled LinwoodChinaMainland feature flag for IntelligenceFlow, which is read by siriAIInChinaMainland.
The flag is currently necessary but not sufficient for Siri AI eligibility on China SKUs.
iOS 27.3 brings Apple Intelligence features to China 9to5mac.com/2026/11/03/ios-2… by @iryantldr
Fun fact: the Dynamic Island banner you see when switching between listening modes on AirPods is a Mac exclusive.
It was already built into macOS 27.0, well before Apple announced the new MacBook Pro.
(Video ft. Dynamic Island on a Studio Display. CLI coming soon.)
macOS 27 paves the way for a touchscreen Mac and a Dynamic Island with the new MenuBarAgent. Some system compact and minimal views already ship.
Siri AI is set to take over the new system aperture with a droplet-like design. Should be a fun day tomorrow!
ALT Five compact system views in macOS 27, rendered in preview pills: AirPods connection, Focus mode activation, AirPods noise cancellation, low battery, and automatic audio switching. Each view has an icon or text label on the left and a status or control on the right.
There’s more: 27.3 introduces a memory graph.
It’s centered on you: relationships, places, routines, health, workouts, and what motivates you or holds you back.
Plus “memory atoms”: facts, opinions, intentions and events, with time, place, people and topic.
Tidbits:
• The memory graph is filled by two kinds of recipes:
1. Fastpass: Turns already-structured data into insights. For example, Person-v1 reads your Me card’s name fields.
2. Agent: An on-device or PCC model extracts insights by querying Siri chats, messages, photos, health data, and other sources.
• There’s only one agent recipe today, MotivatorBarrierResource. Its prompt asks 60 built-in questions, including “Do extended solo media activities during evening downtime routines highlight time availability barriers for adequate sleep?”
• The memory graph is still being built. For example, there’s no process that creates memory atoms yet.
• Memory atoms can generalize into broader memories. Each is tagged with one of 21 topics, such as food, travel, finance, or services, which can store your dentist or lawyer.
Here’s where to find the DBs on Mac:
• Memory graph: ~/Library/CorePersonalization/insight/MemoryGraph/memorygraph.db
• HybridSearch (incl. explicit-memory records, none yet): ~/Library/Daemon Containers/<UUID>/Data/Library/HybridSearch/index/C/index.db
Both use HybridDatabase, Apple’s fork of Kùzu (storage v57), so the public kuzu package can’t open them. You can use HybridDatabase.framework.
Apple is building an explicit-memory store for Siri AI. Memory records are queried via full-text (BM25) and vector search, then merged via RRF.
Siri’s chat recall in 27.2 uses the same search pipeline to find past chats.
Built on the Kuzu graph database Apple acquired.
ALT A Siri AI app icon on a dark gray rounded square, showing a sphere that shades from dark gray at the top to light gray at the bottom, with a ribbon of iridescent light curving across its middle.